跳到正文
arXiv cs.AI· Nadine Chang, Maying Shen, Shizhe Diao, Jialiang Wang, Jingde Chen, Thomas Breuel, Pavlo Molchanov, Rafid Mahmood, Jose M. Alvarez·· 4 小时前AI 评分29

TTCov:面向部署感知学习的测试条件数据筛选方法

Test-Time Coverage: Test-Conditioned Data Curation for Deployment-Aware Learning

AI 导读

TTCov(Test-Time Coverage)是一种数据级测试条件筛选方法,在训练前利用测试侧信息而非在推理时更新模型权重。它通过构建任务 Atlas 与 Knowledge Atlas(K-Atlas)将部署分布转化为筛选目标,再选出部署 AP 分布接近该目标的预算内训练集。

正文 · 原文

View PDF HTML (experimental)

Abstract:Deployed AI systems are often trained from broad candidate data pools, necessitating data curation towards the deployment test distribution. However, standard data curation methods score training-side criteria rather than directly optimizing deployment match. We introduce TTCov (Test-Time Coverage), a data-level test-conditioned curation method that uses test-side information before training instead of updating model weights at inference. TTCov decomposes deployment-conditioned curation into coverage and distribution. To represent coverage, it builds a task Atlas, a collection of LLM-based atomic propositions (APs) describing deployment-relevant concepts, seeded from open task knowledge and expanded with unmatched APs extracted from unlabeled deployment samples. To represent distribution, it instantiates the matched deployment APs with their frequencies, yielding a Knowledge Atlas (K-Atlas) that operationalizes the deployment distribution as a curation target. TTCov then selects a budgeted training set whose deployment APs distribution approximates this target. We apply TTCov towards autonomous driving (AD), keeping adaptation off the inference path while selecting data with greater deployment-relevant coverage, closer K-Atlas matching, and stronger downstream end-to-end driving performance than data-curation baselines, including seamless adaptability to novel domains via city-to-city expansion. Code available at this https URL.
Comments: v2: Added a link to the publicly released codebase
Subjects: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Cite as: arXiv:2607.22697 [cs.AI]
  (or arXiv:2607.22697v2 [cs.AI] for this version)
  https://doi.org/10.48550/arXiv.2607.22697

arXiv-issued DOI via DataCite

Submission history

From: Maying Shen [view email]
[v1] Sat, 18 Jul 2026 00:28:57 UTC (2,167 KB)
[v2] Mon, 16 Nov 2026 19:46:45 UTC (2,153 KB)

来源:arXiv cs.AI · arxiv.org